Example Code

NIWeek 2019 Deep Learning Object Detection using OpenVINO Inference Engine Demo

Code and Documents

Attachment

Overview

 

This example demonstrates the use of Deep Learning APIs to perform Object Detection using both TensorFlow and OpenVINO Optimized Models. The TensorFlow model is converted to OpenVINO model using Model Optimizer. OpenVINO optimizes the TensorFlow model and provides faster inference showed by the OpenVINO Acceleration indicator.

 

Description

 

This example uses a pre-trained TensorFlow Object detection SSD_Mobilenet1_Coco model that has been fine tuned using IC defect images. It also shows the optimized model using OpenVINO’s Model Importer. The Optimized Model was optimized using the “convert2ir.py” script that ships with Vision Development Module 2019.

Both models are being loaded and ran using the Model Importer polymorphic API to detect the defects in the IC part presented. This example demonstrates the gain in execution time of the model with National Instrument’s Inference Engine using OpenVINO for the optimization.

 

For more details, refer to the “OpenVINO Deep Learning in NI Vision” PDF included in the folder.

 

Hardware and Software Requirements

 

LabVIEW 2018 64-bit and later

Vision Development Module 2019

Windows 10 64-bit with Intel 6th generation processor or Linux RT 64-bit target

 

Steps to Implement or Execute Code

 

  • Extract all three folders and place the OV and TF Model folders at the same level as the "NIWeek Deep Learning IC Demo (1900x1080).vi" from the NIWeek OpenVINO2018 folder.
  • Open the Block Diagram of the “NIWeek Deep Learning IC Demo” VI and configure the Vision Acquisition Express VI to use your camera.
  • Run the VI.
  • Observe the OpenVINO acceleration indicator to see how much performance optimization is gained by using OpenVINO.

 

Additional Information or References

 

Front Panel: 

Front Panel of 2019 NIWeek OpenVINO.PNG

T. Le
Vision Product Support Engineer
National Instruments

Example code from the Example Code Exchange in the NI Community is licensed with the MIT license.

Contributors